Pairwise distances like analysis

Dear all,
Currently, I am struggling with one dataset, in which the same animals were sampled 3 times: at the control diet (baseline), then all animals after diet 1 (tp1), and after diet 2 (tp2). So I don't have a control group at tp1 and tp2. Since I have an animal ID in the metadata, I decided to run something similar to pairwise distances (as I understand this plugin, at least).
So steps I took:

  1. Extracted distances of the same animals between baseline and diet 1 (Group 1).
  2. Extracted distances of the same animals between baseline and diet 2 (Group 2).
  3. Performed Wilcoxon test for dependent samples for each group and adjusted p-values.
  4. Performed the Kruskal-Wallis test between groups 1 and 2 to check for a greater effect.

Now after I played with it enough I wonder if I missed something and if such analysis is appropriate.

Hi @timanix,

I think that's a reasonable approach. You might want to adjust for the order, and check the magnitude of the effect between TP 1 and TP 2, just in case you dont have enough of a washout effect. I'd probably also do a permutative t-test over a KW, because I think it lets you get away with fewer assumptions.


1 Like

Hi @jwdebelius,
Thank you for detailed reply :star_struck:! I will take a look on the permutative t-test then. For me, it is expected that tp2 will have a greater effect than tp1 just by diet composition, but I will need to consult with person who implemented the study to inquire washout info.

1 Like